Stability of 8.0.x

MySQL & MariaDB Percona XtraDB Cluster 8.x
3

Hi @matthewb,
many thanks for helping :slight_smile: .

I have monitoring system that saves changes in Xtradb cluster and I’m hitting both issues:

I have big concurrency in updates of data but I balance it to one node and in application logic we retry on lock timeout or deadlock with linear back off.

I don’t see any oom kills or similar. Configuration it the same but 5.7 works great 8.0.x doesn’t :confused: .

Docker image version: 8.0.23-14.1
Config:

[mysqld]
# Tuning
innodb_buffer_pool_size = 6G
wsrep_slave_threads = 64
innodb_read_io_threads = 32
innodb_write_io_threads = 32
innodb_io_capacity = 6000
innodb_io_capacity_max = 8000
innodb_purge_threads = 8
innodb_page_cleaners = 8
innodb_buffer_pool_instances = 6
wsrep_provider_options = "pc.weight=10; gcache.size=1G;"
wsrep_sync_wait = 7

# Limits
max-connections = 10000
open-files-limit = 65535

Only thing I see in logs is deadlock detected, queries stuck in process list and node get’s thrown out of the cluster after some time.

After hanging I noticed nodes get killed with signal 11 and gcache get’s corrupted.

1 Like